/explore

Click through on any links that interest you or select the planets on the right to continue exploring the Outer Web.
You are here

 www.thanassis.space
Optimizing code for the European Space Agency
| | siboehm.com
How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: a Worklog
3.6 parsecs away

Travel
| | In this post, I'll iteratively optimize an implementation of matrix multiplication written in CUDA.My goal is not to build a cuBLAS replacement, but to deepl...
| | seanzhang.me
A hitchhiker's guide to CUDA programming - Ramblings on Deep Learning
3.7 parsecs away

Travel
| | Walkthrough of writing a SGEMM kernel that achieves 95% of cuBLAS performance
| | v0dro.in
Building a FAST matrix multiplication algorithm. | code, travel and explore.
3.9 parsecs away

Travel
| | I've received an assignment for writing a very fast matrix multiplication code using multithreading, BLISLAB, SIMD, etc. In this post I will document my approach to writing this code. I've made the best effort to optimize the multiplication to the hilt, but if readers find anything amiss please leave a comment and I'll have a look at it ASAP.
| | themaister.net
My personal hell of translating DXIL to SPIR-V - part 3 - Maister's Graphics Adventures
25.9 parsecs away

Travel
| [AI summary] The author discusses the complexities and challenges of implementing D3D12's binding model in vkd3d-proton, highlighting the intricacies of bindless resources, local root signatures, and the transition to SM 6.6. They emphasize the need for careful handling of descriptors, validation, and the impact of hardware differences. The author also mentions the benefits of AMD's GCN architecture for bindless operations and the importance of thorough testing and code generation for different GPU architectures.